Julien Chiquet

Tu es chez toi ? Tu utilises la même machine que d'habitude ? Tu es bien connectée à ton compte github ? -- Julien Chiquet Senior Researcher in Statistics, INRAE...

HI, Just to complete Mahendra's answer: we use the `torch_elbo` function for optimisation, but to represent and send back the final value in `optim_par` we use `torch_vloglik` (variational loglikelihood) which...

Found a fix (need to eval the symbol in the correct environment to get the right formula). A PR is en route. Seems to work without any "as.formula" inside/outside the...

@papayoun could you please confirm that the current master version of PLNmodels fixes your problem ? thank you,

Ok, closing since no reaction of @papayoun

@mahendra-mariadassou DO you think it is stll of interest ? I do think so ;-)

Thanks for the report. I am on it, and it seems that there is a sign error in the objective of the VE-step https://github.com/PLN-team/PLNmodels/blob/0a4d0b14e2eb84cd2e429bc8c5419e05f24ce094/src/optim_full_cov.cpp#L169 Since we are optimizing something related...

Ok, now the following piece of code ```r library("PLNmodels") # choose data n

Potentially available in ZIPLN (not yet accessible to the user, only sparsity on the precision matrix is available yet).

Hi, I've made a fix for [quadrupen](https://github.com/jchiquet/quadrupenCRAN), and the package's master branch version is ready to be submitted to CRAN (it passes the checks on the github and winbuilder runners)....